Adaptive Classification for Concept Drifting Data Streams with Unlabeled Data

Abstract

The research in data stream mining has gained a high attraction due to the importance of its applications and the increasing generation of streaming information. The data streams are the set of data, which are moving in specified distribution. Any changes or variation in their target can be considered as concept drift. Recent researches are concentrating reducing the level of concept drift. In this paper, we have planned to propose a concept drift controlling in unlabeled data streams. we have planned to incorporate a clustering technique and the decision tree algorithm to control the concept drift in the unlabeled data. We have planned to start the training process with the unlabeled data. After the training process the data are labelled and a decision tree is created for the train data. This data is again tested for error value for a particular number of iteration. The experimentation is conducted by using sky concept dataset. The experimental evaluation produced satisfactory results. The minimum error rate obtained is in the range of 0.02 to 0.3 percentages.